Focus Groups as Qualitative Research

Transcription

1 Focus Groups as Qualitative Research PLANNING AND RESEARCH DESIGN FOR FOCUS GROUPS Contributors: David L. Morgan Print Pub. Date: 1997 Online Pub. Date: Print ISBN: Online ISBN: DOI: / Print pages: This PDF has been generated from. Please note that the pagination of the online version will vary from the pagination of the print book.

2 / n4 PLANNING AND RESEARCH DESIGN FOR FOCUS GROUPS This chapter addresses the planning that must be done prior to doing focus groups. Both it and the next chapter on how to conduct and analyze focus groups will have more emphasis on technical procedures. Such practicalities are an important aspect of any research technique, and focus groups do not differ greatly from other qualitative methods of data collection in this regard. In particular, the framework for the next two chapters is based on Kirk and Miller's (1986) general description of the four phases of qualitative research: planning, observation, analysis, and reporting. Of these four phases in the life of a research project, this presentation will devote one entire chapter to the planning phase. I emphasize planning because this is the area where focus groups depart most from standard practices in other qualitative methods. The fact that they are group interviews is the source of most of these planning needs. In particular, the need to bring together several participants requires attention to who the participants are and how the researcher will interact with them as a group. Once the choices surrounding the format of the group interview are made, the subsequent observation, analysis, and reporting phases mostly pose issues that are already familiar to experienced qualitative researchers. Before Starting Three obvious factors that affect the ability to plan are ethical concerns, budget issues, and time constraints. In many respects, the ethical concerns in focus groups are similar to those raised in all qualitative research (Punch, 1986), but the specific concerns generated by focus groups also require attention (Smith, 1995). Issues concerning invasion of privacy are especially important whenever taping is the primary means of data collection. Actual audio and visual presentations of tapes are relatively rare in the social sciences, but they can be very tempting in the case of focus groups. No amount Page 2 of 18

3 of accuracy in transcription will ever substitute for the excitement of actually listening to an emotional exchange among participants, and videotapes can be even more seductive. It is thus wise to decide up front who will hear or see the tapes. My advice would be to limit this access to the research staff; unless you know from the beginning that public presentation of the tapes will be an integral part of the research, it is best to rule out this option altogether. [p. 32 ] One unique ethical issue in focus groups is the fact that what participants tell the researcher is inherently shared with other group participants as well. This raises serious invasion of privacy concerns and effectively limits the kinds of topics that the researcher can pursue. Such limitations are actually practical as well as ethical: It is not a productive use of focus groups to ask people to talk with discussion partners with whom they are not comfortable. In other words, most focus groups that involve an invasion of privacy are also a waste of the researcher's time. Note, however, that this argument does not necessarily apply to groups of self-acknowledged deviants, especially if they are members of deviant subcultures in which the informal equivalent of focus group discussions occur. That case requires the usual protection of participants, with the added assurance that all the participants in each discussion truly belong to the shared milieu. Turning to budget issues as a factor in planning focus groups, one guideline is that marketing researchers often charge clients over $2,000 per group when participants come from the general population, and this figure may double when working with specialized groups of participants who are more difficult to recruit. If the research is being subcontracted to a marketing firm, this is what it will cost. If you are doing some of the work yourself, it will probably cost substantially less. Major cost factors include salaries to moderators, travel to research sites, rental of research sites, payments to participants, and producing and transcribing tapes. Many of these costs are essentially fixed by the circumstances of the specific research project, but substantial savings are possible if the researcher has the time and skills necessary to perform the moderator function. This is especially true if the only alternative is hiring an outside moderator because these skills often have a price tag that is more Page 3 of 18

4 in line with the resources of corporate clients rather than academics. One cost-saving compromise is to hire a moderator whose sole role is to conduct the groups and not to analyze the data. This not only takes advantage of the professional moderator's specific skills but also recognizes that the task of making sense of the data is best handled by those who will ultimately be responsible for its use. Even this compromise, however, requires that the research team allocate both time and money for working with the outside moderator to plan the project because both parties must have a truly shared understanding of the research objectives and procedures. Social science research projects vary considerably in the time constraints that they place on focus groups. When the focus groups serve a specific [p. 33 ] purpose within a larger project, such as producing questions for a survey instrument or guiding the design of an intervention program, then such groups may need to be done on tight timelines. Typically, the time savings for such work comes from less detailed analyses; because the uses for the groups are quite explicit, the summary of their content is more straightforward. In these limited cases, the entire focus group project may be completed in just a few weeks. The kind of qualitative research that would satisfy journal reviewers or a dissertation committee, however, may require time for not only more detailed analyses of transcripts but also the possibility of returning to the field for more data. In the planning phase, the design decisions themselves typically require at least 2 weeks more if the project is complex or if the research team is inexperienced with the method. Furthermore, the recruitment of participants may be quite time-consuming, especially when using specialized populations or comparing multiple categories of participants. In the observation phase, although each group takes only a few hours, conducting two a day, or even five a week, can be a killing pace without either a very large staff or considerable reliance on outside services. Finally, in terms of analysis and reporting, transcript typing is slow, and transcript analysis is very time-consuming. Depending on the number of groups, the availability of the participants, and the kind of analysis intended for the transcripts, count on a project taking between 3 and 6 months (longer if the staff divide their time between this project and other commitments). From the beginning, it is important to have realistic expectations not just with regard to budget and time but also in terms of the total amount of investigator effort that is necessary to produce the desired data. This is hardly a new problem, but because of Page 4 of 18

5 the relative novelty of focus groups in social science research it deserves attention. The reputation that focus groups have in some circles as a quick and cheap technique is due to the very limited function to which they have too often been relegated: preliminary explorations to set the stage for real research. Indeed, Morgan and Krueger's (1993) catalog of myths about focus groups includes the idea that they are either quick or cheap. When pursued as a self-contained research technique, focus groups demand the same attention to detail as any other means of data collection. As is always the case, the quality of the data depends on the quality of the preparation: Careful planning cannot guarantee insightful results, but a cavalier approach to the design and execution of the research is almost certain to produce poor results. [p. 34 ] Overview of Focus Group Research Design The planning for a focus group project includes a number of decisions about how the data will be collected. Considered in order of their impact on the nature of the data, the first decision concerns who will participate in the groups. The next decision determines how structured the groups will be, including the level of moderator involvement. After that, there are further decisions about the size of each group and the number of groups in the total project. Over the years, a number of rules of thumb have evolved to capture the most common choices that researchers have made with regard to each of these decisions (Morgan, 1992a). According to these rules of thumb, focus group projects most often (a) use homogeneous strangers as participants, (b) rely on a relatively structured interview with high moderator involvement, (c) have 6 to 10 participants per group, and (d) have a total of three to five groups per project. Unfortunately, some people act as if these rules of thumb constitute a standard about how focus groups should be done rather than a descriptive summary of how they often are done. In reality, most projects have some elements that require special attention, and it may be relatively rare for a project to match all four of these criteria. Page 5 of 18

6 These rules of thumb are most useful as a point of departure in the planning process. With regard to the choice of who the participants will be, it could be useful to ask whether homogeneous strangers would best serve the purposes of this particular project. This leads directly to questions about whether mixed groups would be more productive than homogeneous groups for this research topic or whether it is even realistic to try to recruit groups of strangers in this particular research setting. A similar self-questioning would apply to each of the other rules of thumb: Would a less structured interview or a lower level of moderator involvement produce more productive discussions on this topic among these participants? Would smaller groups produce more detailed data from each participant or would larger groups produce a wider range of ideas? Would it take a larger number groups to cover the participants' range of experiences and opinions on this topic or are there reasons to believe that a smaller number of groups will be sufficient? For any given decision, this self-questioning may result in staying with the original rule of thumb. If these rules of thumb do indeed represent the most frequent choices, then this is hardly surprising. Even so, a choice to go with a rule of thumb still needs to be a careful decision that [p. 35 ] is based on the needs of the specific project. The remaining sections of this chapter will provide guidance on making such decisions. Determining the Types of Participants SAMPLING CONCERNS In selecting participants for a focus group project, it is often more useful to think in terms of minimizing sample bias rather than achieving generalizability. Focus groups are frequently conducted with purposively selected samples in which the participants are recruited from a limited number of sources (often only one). Such bias is a problem only if ignored that is, interpreting data from a limited sample as representing a full spectrum of experiences and opinions. If a particular recruitment source does limit the nature of the data that are available, then this forces the choice between living with those limitations or finding other sources of participants that will reduce these biases. Page 6 of 18

7 The shift away from an emphasis on generalizability also means a shift from random sampling toward theoretically motivated sampling. Random sampling is seldom of use in selecting participants for focus groups for at least two reasons. First, the small number of participants involved in most focus group projects makes it extremely unlikely that a sample of size 40 or so will be adequate to represent a larger population, regardless of random selection. Second, a randomly sampled group is unlikely to hold a shared perspective on the research topic and may not even be able to generate meaningful discussions. In contrast, the following section presents several reasons for selecting focus group participants through purposive or theoretical sampling (Glaser & Strauss, 1967; Patton, 1990). HOMOGENEITY AND SEGMENTATION The decision to control the group composition to match carefully chosen categories of participants is known as segmentation. Segmented samples are closely tied to the emphasis on homogeneity in the composition of focus groups. It is this homogeneity that not only allows for more free-flowing conversations among participants within groups but also facilitates analyses that examine differences in perspective between groups. For example, if sex differences affected either the participants' comfort in the discussion or the analyst's ability to make useful comparisons, then there would be [p. 36 ] advantages to conducting separate groups of men and women that is, segmenting by sex. When are segmented samples and homogeneous groups the most appropriate choice? The group composition should ensure that the participants in each group both have something to say about the topic and feel comfortable saying it to each other. Try asking whether these participants could easily discuss this topic in normal, day-to-day interaction. Participants must feel able to talk to each other, and wide gaps in social background or lifestyle can defeat this requirement. Note, however, that the goal is homogeneity in background and not homogeneity in attitudes. If all the participants share virtually identical perspectives on a topic, this can lead to a flat, unproductive discussion. Page 7 of 18

8 The most common background variables that are considered in running mixed versus segmented groups are sex, race, age, and social class. Whether the sexes interact differently in mixed groups is a longstanding research question (Thorne & Henley, 1975); therefore, that alone may convince some researchers to segment by sex. This concern is most acute, however, when the issues raised by a given topic correspond to differences in perspectives between men and women. Such differences in perspectives may either reduce the comfort level in the discussion or affect how clearly either perspective gets discussed. Similar remarks apply to race, although, given the rather selective integration of American society, there may be even more topics in which racial differences in perspective could become an issue during group discussions. Older and younger participants may also have difficulty communicating with each other either because they have different experiences with a topic or because similar experiences are filtered through different generational perspectives. Class differences reflect a general segregation of interaction in our society so that even when the participants have few overt class differences in their experiences they may still be uncomfortable discussing personal experiences in each other's presence. This last point illustrates a more general concern: It is not the actual differences among participants but whether they perceive each other to be different that determines their willingness to discuss a topic together. I illustrate these points using my research on widowhood (Morgan, 1989), in which most of the participants were women in their sixties and seventies, but one group had two male participants and another group contained a very young widow. In both these cases, there was a moderate disruption in the flow of the discussion because the other participants went out of their way to be solicitous to the outsiders in their group. All the groups, however, were quite mixed with regard to social class and this had [p. 37 ] little noticeable impact because these widows explicitly maintained that their bereavement created a fundamental similarity that overshadowed the differences in their backgrounds. In other words, their shared beliefs determined what made another's experience similar or different. The choice between mixing and separating categories of participants also occurs when the participants occupy different social roles with regard to a topic. For example, the difference between fathers and mothers in a discussion of child-rearing practices is not just a difference between men and women. This issue is particularly common Page 8 of 18

9 in organizations, in which individuals in various positions have routine patterns of what they do and do not discuss together. Differences in authority or status are particularly likely to create this problem, and such differences may occur in the general community as well as in organizations. There are strong arguments (Morgan & Krueger, 1993) against mixing categories of participants across authority or status lines, either due to ethical issues or because of the high probability that the discussion will be uncomfortable at best and conflict-ridden at worst. Using groups that are segmented by background or role-based differences has the cost of requiring more groups because it takes a certain minimum number of groups within each category to observe that category's range of responses to a topic. Using multiple segmentation criteria makes the decisions about group composition more like an experimental design. Knodel (1993) describes a useful example of such a study. The general strategy in using complex, segmented designs is to create a variety of internally homogeneous groups that capture a wide range of potentially distinct perspectives (Kitzinger, 1994a). Unfortunately, this approach can also make the data collection quite expensive and the analysis quite complex. One way to address this issue during the planning phase of the research is to try one pretest group that is mixed and then compare it to other groups that separate the categories in question. Comparing these discussions should reveal the comfort level in mixed groups. STRANGERS VERSUS ACQUAINTANCES A final decision in determining the group composition involves seeking out strangers versus allowing acquaintances to participate together. The rule of thumb favors strangers because, although acquaintances can converse more readily, this is often due to their ability to rely on the kind of taken-for-granted assumptions that are exactly what the researcher is trying to investigate (Agar & MacDonald, 1995). This problem is even more [p. 38 ] severe when the assumptions among acquaintances include invisible boundaries around the subjects that they have tacitly agreed not to discuss. Page 9 of 18

10 The notion that focus groups must consist of strangers, however, is certainly a myth (Morgan & Krueger, 1993). In fact, social scientists routinely conduct focus groups in organizations and other naturally occurring groups in which acquaintanceship is unavoidable. Furthermore, working with prior acquaintances can help the researcher deal with issues of self-disclosure (Jarrett, 1993). The real issue is that strangers and acquaintances can generate different group dynamics, which may lead a researcher to different choices, depending on the nature of the research goals. Where differences in group dynamics are not an issue, practical concerns may govern the choice between strangers and acquaintances. In some cases, it may be almost impossible to recruit a full group of acquaintances (e.g., among service recipients); in other cases, it may be almost impossible to avoid it (e.g., in organizational settings). For these circumstances, decisions should rely on the basic criterion of whether a particular group of participants can comfortably discuss the topic in ways that are useful to the researcher. RECRUITMENT ISSUES To this point, this discussion has concentrated on determining the content of the sample, but an equally important set of issues concerns the recruitment of that sample. I have even argued that inadequate recruitment efforts are the single most common source of problems in focus group research projects (Morgan, 1995). Simply locating participants and getting them to agree to show up is often not enough; instead, it is essential to develop careful procedures that ensure that enough participants actually do show up for each group. Projects that call for specific categories of participants require special recruitment efforts. Telephone screening interviews are one useful approach. This involves making phone calls, through either random-digit dialing or a predetermined list, and then asking a very short questionnaire to see if there is anyone in the household who both fits the recruitment category and is interested in participating. Such screening is more than just a mechanism for locating participants because homogeneity may be vital for the group's ability to share a discussion of the research topic. Even if only one of the participants fails to share some crucial characteristic, the discussion can get totally off Page 10 of 18

11 track. If homogeneity is a major concern, then a further screening questionnaire can be administered to potential participants [p. 39 ] as they arrive for the discussion. Be warned, however, that screening questionnaires do run the risk of alerting participants to the topic of the research or at least creating such expectations, whether they are accurate or not. When working with highly specialized categories of participants, recruitment procedures have to be equally specialized. In particular, it may be necessary to use substantial cash incentives to recruit top-level executives or others with unusual expertise; payments of $100 per person in such sessions are not uncommon. In community samples, marketers typically pay participants $25 to $50. Fortunately, there are many nonmonetary incentives for participating in focus groups (Krueger, 1994). In particular, if the research has an external sponsor that is meaningful to the participants (such as a popular community organization), involvement with this sponsor may substitute for a cash incentive. In dealing with uncertainty about incentives or other difficulties in the recruitment process, one general strategy is to conduct a few key informant interviews on the subject of recruitment even qualitative researchers occasionally need to be reminded of the value of getting their participants' perspectives on a problem. Determining the Level of Group Structure Choices about interview standardization and moderator involvement go together to determine how structured the group discussion will be. Interview standardization refers to whether the same questions are asked of every group that is, the extent to which the content of the interview is either predetermined or flexible. Moderator involvement refers to the management of the group dynamics that is, the extent to which the moderator either controls the discussion or allows relatively free participation. Although there are various strategies for combining different degrees of interview standardization and moderator involvement (Morgan, 1992a), most projects tend to set both of them at comparable levels, which can be referred to as more structured or less structured approaches. Page 11 of 18

12 MORE STRUCTURED GROUPS More structured approaches to focus groups are especially useful when there is a strong, preexisting agenda for the research. Both a standardized interview and a higher level of moderator involvement contribute to meeting this set agenda. When the project begins with a strong sense of what the research questions are, then a standardized interview will make sure that all the groups discuss these issues in a relatively comparable [p. 40 ] fashion. In addition, a higher level of moderator involvement will keep the discussion concentrated on the topics that interest the researchers rather than extraneous issues. Examples of goals that often rely on more structured approaches include providing inputs to other research efforts, such as program designs or survey content; making consistent comparisons across all the groups in a set; or comparing the thinking of a new set of participants with a previous set of groups. The most obvious problem with more structured approaches is that a narrow set of questions or a motivated moderator may well produce equivalently limited data. Worse, it may be difficult to know that this is the case. Once participants sense that there is a distinct agenda for the discussion and that the moderator is there to enforce that agenda, then they are likely to acquiesce in all but the most extreme circumstances. More structured approaches thus pose a trade-off between more ability to hear about what interests the researchers and less ability to be sure that this is what actually matters to the participants themselves. LESS STRUCTURED GROUPS Less structured approaches to focus groups are especially useful for exploratory research. When the basic issues are poorly understood or existing knowledge is based on researcher-imposed agendas, then an unstandardized interview guide will provide the opportunity to hear the interests of the participants themselves in each group. In addition, minimizing the moderator's involvement in the discussion will give the participants more opportunity to pursue what interests them. What makes less Page 12 of 18

13 structured focus groups such a strong tool for exploratory research is the fact that a group of interested participants can spark a lively discussion among themselves without much guidance from either the researcher's questions or the moderator's direction. In other words, if the goal is to learn something new from the participants, then it is best to let them speak for themselves. A major disadvantage of less structured groups is that they are more difficult to compare from group to group. In particular, topics will come up in some groups and not in others. This difference in the topics that are raised from group to group makes the data more difficult to analyze than the well-ordered discussions that more structured approaches produce. The trade-off with less structured approaches is thus between a greater ability to learn about the participants' perspectives in their own words and less ability to pursue any aspect of these perspectives in a consistent fashion. [p. 41 ] I illustrate the different uses for more structured and less structured approaches through a comparison of two studies I did with people who were caring for a family member with Alzheimer's disease (Morgan, 1992b). In one study, I used a more structured approach to compare the decision-making processes of caregivers who had brought their family members in for a diagnosis when the symptoms were either relatively mild or more severe. The goal in that study was to pinpoint the influence of several well-known factors in the caregivers' decision-making processes. To accomplish this, a structured interview guide walked the caregivers through the history of their decision making, and the moderator controlled the discussion to be sure that every caregiver had a roughly equal chance to tell his or her story. In the other study, I used a less structured approach to compare caregivers who had placed their family member in an institution, such as a nursing home, with those who were still providing care in the community. The goals in that study were exploratory because relatively little was known about family caregiving in nursing homes; therefore, we wanted to compare the perspectives of community-based and nursing home-based caregivers. The interview guide in that study simply asked to hear as much as possible about what made caregiving either easier or harder, and the moderators minimized their direct involvement in the groups. Page 13 of 18

14 THE FUNNEL AS A COMPROMISE APPROACH To this point, I have contrasted more structured and less structured approaches to focus groups as a way to emphasize the need to make research design choices about standardized versus unstandardized interview guides and high versus low levels of moderator involvement. It is also possible, however, to design a compromise between the two by using what is known as a funnel strategy. In a funnel-based interview, each group begins with a less structured approach that emphasizes free discussion and then moves toward a more structured discussion of specific questions. The funnel analogy matches an interview with a broad, open beginning and a narrower, more tightly controlled ending. This compromise makes it possible to hear the participants' own perspectives in the early part of each discussion as well as their responses to the researcher's specific interests in the later part of the discussion. Although a funnel seems to offer the best of both worlds, it can be harder to achieve than either a more structured or a less structured approach. Blending the unstandardized opening with later, fixed questions requires [p. 42 ] care to find both a good starting point and a successful transition to the more controlled set of topics. Furthermore, shifting from a less involved to a more involved moderator style requires avoiding either an overly directive moderator style in the beginning or a nondirective style later on. Funnel designs thus work best in projects that truly need a combination of more structured and less structured data, and they are unlikely to save a project in which the researchers are simply uncertain about what kind of data they need. Determining the Size of Groups The amount that each participant has to contribute to the group is a major factor in decisions about group size. If the participants have a low level of involvement with the topic, it may be difficult to maintain an active discussion in a smaller group. Another key factor is how much detail the researchers need to hear from each participant. Small groups also run the risk of being less productive because they are so sensitive to the Page 14 of 18

15 dynamics among the individual participants. In particular, the functioning of the group as a whole can easily be disrupted by friendship pairs, experts, or uncooperative participants. Small groups thus work best when the participants are likely to be both interested in the topic and respectful of each other. In addition, small groups are more useful when the researcher desires a clear sense of each participant's reaction to a topic simply because they give each participant more time to talk. Larger groups have a different set of problems that may limit their productivity. In particular, it is typically more difficult to manage their discussions, and this is especially true when the participants are highly involved in the topic. In practice, large groups can easily break up into small conversations among neighbors around the table, or people may start talking at once; either of these problems implies a loss of data because such conversations are very difficult to tape. Consequently, large groups typically require a higher level of moderator involvement, and it takes an experienced moderator to control them without engaging in continual efforts at discipline. As previously noted, however, such high levels of moderator involvement are not desirable for some research purposes. One practical consideration is that, whatever size is selected, it is important to overrecruit to cover for no-shows. The common rule of thumb is to over-recruit by 20%, although the actual extent of over-recruitment depends on who the participants are, whether they are being paid for their [p. 43 ] participation, where the groups are conducted, and how vital the desired group size is for the overall design of the research. Combining both practical and substantive considerations helps to clarify the basis for the rule of thumb size that specifies a range of 6 to 10. Below 6, it may be difficult to sustain a discussion; above 10, it may be difficult to control one. Even within this range, choices may still be necessary. For some topics, such as those in which participants have either relatively high or relatively low levels of involvement, there could be a world of difference in the dynamics for a group of 6 versus a group of 10. Also, one should not feel imprisoned by either this lower or upper boundary. I have conducted groups of 3 highly involved participants that would have been unmanageable at size 6, and I have led discussions in naturally occurring groups of 15 to 20 in which the process was quite orderly. Ultimately, both the purposes of the research and the constraints of the field situation must be taken into account. Page 15 of 18

16 Determining the Number of Groups Although I have held off discussing the number of groups until last, this is not because this topic is less important than the others. In particular, the number of groups in the project will be the primary determinant of how much data the research produces. The number of groups also has a direct impact on the size and structure of the research team. Conducting many groups almost ensures the need for a larger research staff, the only other alternative being to extend the data collection and analysis over a longer period of time. The basis for the rule of thumb that projects should consist of three to five groups comes from a claim that more groups seldom provide meaningful new insights. In both the social sciences (e.g., Zeller, 1993) and marketing (e.g., Calder, 1977), this is frequently summarized as the ability to stop collecting data when the moderator can accurately anticipate what will be said next in a group. Seasoned qualitative researchers will recognize this as another way of expressing the goal of saturation (Glaser & Strauss, 1967) that is, the point at which additional data collection no longer generates new understanding. Whether three to five groups will be adequate for saturation, however, is an issue that depends on many factors. Probably the most important determinant of the number of groups is the variability of the participants both within and across groups. Within groups, projects that bring together [p. 44 ] more heterogeneous participants will typically need more total groups because the diversity in the group often makes it more difficult to sort out coherent sets of opinions and experiences. Across groups, projects that compare several distinct population segments will typically require more total groups to achieve saturation within each segment. Another factor that affects the number of groups is the degree of structure in the interviews. In general, projects that use less standardized interviews and lower levels of moderator involvement require more groups. The reason is that both of these factors increase the variability from group to group. Page 16 of 18

17 A different set of issues concerns the sheer availability of participants. If there are few potential participants available or if they are highly dispersed, then it is wisest to run several groups of smaller size. This is implicit in the criterion of saturation because it is necessary to compare the discussions from several groups to determine whether the participants are repeating what was said in earlier groups. Regardless of the circumstances, collecting only one group creates severe problems. The problem with having only one group is that it is impossible to tell when the discussion reflects either the unusual composition of that group or the dynamics of that unique set of participants. Even when there are data from just two groups, if what they say is highly similar then this provides much safer ground in concluding that group dynamics were not responsible for this content. Also, if the discussions in the two groups differ, then this is a fair warning that saturation has not been achieved. A different version of the same problem occurs in projects that use multiple segments because the substantive content of the group's discussion is confounded with its unique composition and dynamics. There should thus be more than one group in each segment, which will obviously increase the total number of groups in groups that use multiple segments. In general, the goal is to do only as many groups as are required to provide a trustworthy answer to the research question due to the costs involved in conducting more groups. These costs involve not only additional efforts in recruitment and data collection but also additional coding and analysis efforts with a larger number of transcripts. There are dangers in plans that call for using a bare minimum number of groups, however, because this can make the project as a whole vulnerable when a single group fails (e.g., because not enough participants show up or because a group is so unusual as to be of dubious utility). The safest advice is to determine a target number of groups in the planning stage but to have a flexible alternative available if more groups are needed. [p. 45 ] Page 17 of 18

18 Summing Up Clearly, there are a great many issues to be considered in planning for focus groups. This chapter has only touched on a set of general issues that are likely to affect nearly all focus groups; any specific research project will confront many more issues than these. In particular, several of the issues covered in the following chapter have additional planning implications. Just as important, it would be unfortunate to give the impression that planning is a discrete stage in the research project. Planning is not something that is over and done by a given point, with a totally new set of concerns waiting to replace it. Even when the goal is to do a quick and inexpensive set of focus groups, this does not eliminate the need for planning. Indeed, this chapter should have made it painfully obvious that the only way to take advantage of the flexibility of focus groups is through a diligent effort at prior planning! / n4 Page 18 of 18